Then I could finally finish the last part of OCXO assembly by installing the LM1085. Little did I know that the OCXO would fail miserably...
When I slowly increased the voltage to the 004 equivalent assembly, I noticed that the power supply seemed to be faulty. The voltage didn't go up as much as expected. But the amps did, rising to ~3A. And I could tell that the heatsink I had made was getting really warm. So I measured the resistance between the assembly voltage input pin and GND, and it was above 100 Ohms and not very low/shorted. However capacitor C105 was shorted. And so were the other capacitors between the 5V rail (LM1085 OUT) and GND, C104, C106 and C107. Then I looked at the LM1085 datasheet and noticed that the mounting tab, where the 3mm mounting bolt goes through, was not GND, but voltage output. Therefore I removed the bolt going through the LM1085, the heatsink and the PCB. I was not convinced this would make a difference, as the PCB ground plane was not exposed around the hole for the LM1085 on either side. And everyone else that have built his thing have just put a metal bolt straight through. Nevertheless, the OCXO assembly was now reacting very differently as I applied power to it. I could see the 10 MHz output on the scope as well, as I noticed that the current, that this time had started out at ~0.6A, was decreasing. Success!
With the assembly running, I reinserted the 3mm bolt, and all was OK initially. I had to tightened the nut a little bit before the output of the OCXO was affected. Maybe the designer, amc184, or someone else that can read the gerber design files in reply #12 in this thread can figure out if the bolt hole is drilled through the copper ground plane of the PCB. Maybe it barely makes contact with the ground plane edge in the hole as the nut is tightened?
Now I only have to find a way to fasten the LM1085 to the rest of the assembly. It is odd that everyone else that have done this job before me seem to have had success by simply putting a plain bolt straight through, without electrically insulating it.
By the way: the 20 turn trimpot is a very good design choice (made by
amc184) to adjust the frequency output of the OCXO. No need to have the right tongue angle, as one full turn makes less than 0.1 ppm (1Hz) change.